04:50marcus: what exactly are the maxwell changes on nouveau from kernel 4.0 to 4.1rc? has basic exa support been introduced already?
05:54kisisten: imirkin_: I've tried all the opengl compositors with my nvidia, but none of them are working, they all freeze when I connect the external vga monitor
05:54kisisten: only xrender works
08:03Hedgework: Hey, all. I'm setting up a new Gentoo system on a laptop with a Kepler card (Nvidia Quadro K2100M) and Xorg and I aren't getting along...
08:04Hedgework: When I try to start X via "startx" as a regular user, xorg-server crashes, and my laptop's screen turns off (can't seem to get it back on without a reboot).
08:04Hedgework: The machine is still responsive over ssh, so I was able to confirm that xorg-server wasn't running, etc.
08:05Hedgework: Output of dmesg at https://bpaste.net/show/5573cfb5556e
08:05Hedgework: Xorg.0.log at https://bpaste.net/show/cf8eff2ac294
08:05Hedgework: Any help would be appreciated. None of my other machines have Nvidia chipsets, so I may have missed something very basic in setting up nouveau.
08:26night199uk: anyone know if there is a better reference of the init script table opcodes than nouveau source?
08:26night199uk: guess they’re not documented anywhere
08:27night199uk: i see some code with opcode 0x7c, which doesn’t seem to be implemented by nouveau
09:55imirkin_: Hedgework: you might just be SoL with that one... is this a laptop?
09:55Hedgework: imirkin_: Yes.
09:56Hedgework: imirkin_: The one work got me, so I was only semi-involved in its selection. Happy so far apart from that I need to run X for a few things. There MUST be a way for it to work with nouveau, because it does so when I boot with sysrescuecd and start X there.
09:56imirkin_: Hedgework: sure, just boot with nouveau.noaccel=1
09:57imirkin_: this should get you fully functional modesetting and no 2d/3d accel
09:57Hedgework: "boot with" as in...? Kernel param?
09:57imirkin_: or module param
09:57Hedgework: imirkin_: I'm not gaming, I'm doing infosec...I don't care about accel.
09:57Hedgework: will try, thanks.
09:57imirkin_: you say that now
09:57imirkin_: but you probably use a desktop
09:58imirkin_: which people have decided would be a GREAT idea to do full-screen pixel shaders on
09:58imirkin_: so now those will be running with llvmpipe
09:58imirkin_: which is great, but CPU's aren't as good at this sort of thing as GPU's
09:58Hedgework: imirkin_: 90% of what I do is command-line. I only use gui for web browsing and coping with the godsawful malformed-courtesy-of-MS-Outlook-web-client email certain people insist on sending me.
09:59imirkin_: just make sure to disable any GL compositors your desktop might default with, and you should be perfectly fine
09:59imirkin_: you should still be able to use multiple screens, etc
09:59Hedgework: I use i3wm, there are no compositors in the default (or in my) config...not even sure if it's possible.
09:59imirkin_: you should be all set then ;)
10:00imirkin_: someone's working on figuring out why the laptop GK106's don't work, but i think it's slow-going
10:00imirkin_: i think indications are that it's some timing issue
10:01Hedgework: Thanks for the info, imirkin_.
10:03imirkin_: np. good luck. let me know if noaccel doesn't help and i cna look further
10:03imirkin_: i can never remember *exactly* what the various options do
10:05Hedgework: will do...I need to get to a stopping point before I can reboot and try it
10:36imirkin_: skeggsb: take a look at Hedgework's dmesg... it looks like his PDAEMON is dead -- reading PDAEMON.PERIODIC_ENABLE generates a mmio fault
10:37imirkin_: Hedgework: mind mentioning what laptop model you have?
10:37glennk: one of those hybrid laptops?
10:37imirkin_: ah nm. dmi to the rescue. DMI: TOSHIBA TECRA W50-A/TECRA W50-A
10:37imirkin_: from the looks of it, it's actually only the nvidia gpu exposed
10:38Hedgework: imirkin_: Toshiba Tecra...I can look up the model # in a sec.
10:38Hedgework: imirkin_: there you go :)
10:38Hedgework: imirkin_: correct, lspci can't see any other video chipset.
10:38glennk: should be a bios option to disable the dgpu?
10:38Hedgework:already thought of just reverting to crappy onboard video if it were there.
10:38Hedgework: glennk: there does not seem to be.
10:39imirkin_: i assume you poked around in the bios?
10:40glennk: intel hd 4600, should be not too awful
10:40imirkin_: glennk: sometimes it's not "pinned out", so to speak
10:40imirkin_: esp on dumb "high end" laptops
10:40glennk: oh right, toshiba...
10:41imirkin_: yes, the leader in dumb :)
10:41Hedgework: imirkin_: of course.
10:41Hedgework: The firmware on this laptop is atrocious.
10:41Hedgework: But, it was the best of the available options given university purchasing rules (which are braindead).
10:41imirkin_: long gone are the days whene everyone had shared identically horrible firmware. now everyone has differently horrible firmware.
10:42glennk: eh, ibm etc have always had their own horrible firmware
10:42Hedgework: Basically, it was this, a macbook pro, a useless netbook/chromebook, or one of a number of crappy Dells (all of which have heat problems).
10:43imirkin_: for all their faults, lenovo has thankfully not killed off the ibm keyboards. which is my #1 criteria when selecting laptops
10:45Nebukadneza: and clit-mouse!
10:45imirkin_: i prefer 'nipple', but sure.
10:45glennk: i always stab myself on the nipple
10:46glennk: but i think you're right imirkin_, this toshiba appears to have the display connected to the nvidia gpu
10:46Nebukadneza: it’s the number one reason for thinkpads…
10:47Hedgework: imirkin_: Yes, they have...all the new thinkpads are chiclet keyboards.
10:47Hedgework: imirkin_: My personal laptop is a Thinkpad, but old enough to have a decent keyboard.
10:48imirkin_: oh really? haven't bought one in a while. but the one i have (T420s or so) is just fine
10:48imirkin_: i heard that the travel is considerably reduced on the X1 Carbon
10:48Nebukadneza: really all? my last info was that you need to pay extra for one…
10:49imirkin_: but that it was still pretty good
10:49Hedgework: Nebukadneza: If there's an option to get a real keyboard, I did not find it while shopping their site.
10:49Hedgework: I ended up repairing my T420 instead of replacing it.
10:49Karlton: Hedgework: out of all those, the chromebook might of been the best choice(assuming it was one where you could at least replace part of the boot firmware)
10:50Nebukadneza: Hedgework: was a while ago…
10:50Hedgework: Karlton: I've got to be able to run multiple VMs while traveling, and stuff like johntheripper, which is probably beyond the resources of a chromebook.
10:51Karlton: yeah RAM might be the only limiting factor or if it is an older model chromebook
10:52Hedgework: Also storage, by the time I generate some large-ish rainbow tables.
11:12Hedgework: imirkin_: Thanks! I have functioning xorg now on that laptop.
11:13glennk: that was with nouveau.noaccel=1 ?
11:14Hedgework: glennk: Yes.
11:51Hedgework:just wishes this meeting would end so she can finish building her system. :)
13:03ergegegfwrg: [ 7.643356] nouveau [ CLK][0000:01:00.0] 07: core 270 MHz memory 405 MHz [ 7.643364] nouveau [ CLK][0000:01:00.0] 0f: core 775 MHz memory 900 MHz [ 7.643515] nouveau [ CLK][0000:01:00.0] --: core 270 MHz memory 324 MHz
13:03ergegegfwrg: Someone was right.
13:03ergegegfwrg: How do I reclock it?
13:04imirkin_: what gpu is this?
13:04imirkin_: [looks like a fermi?]
13:04imirkin_: i'm looking for a GFxxx or GKxxx or GMxxx
13:05ergegegfwrg: Hm, wait.
13:05imirkin_: should be in the logs. or in lspci -nn -d 10de:
13:06imirkin_: then step 1: implement reclocking for fermi :)
13:06ergegegfwrg: But it sets clock at boot.
13:06imirkin_: the vbios does, yea
13:07ergegegfwrg: So, there's no any method to clock it right?
13:07imirkin_: install nvidia blob drivers
13:08ergegegfwrg: Will I be able to use it without bumblebee, just offloading?
13:09imirkin_: with blob drivers? no clue.
13:09ergegegfwrg: Okay, thanks.
13:41tobijk: imirkin_: push the phi patch and the op_cvt to master? :)
13:44imirkin_: tobijk: i want to write a piglit that fails with the current phi moves pass logic
13:44imirkin_: but... i'm failing at the failure
13:46tobijk: ah allright
13:46imirkin_: which is making me reconsider the fix in the first place... perhaps that handleTXL thing is what actually need fixing
13:47imirkin_: although the PhiMovesPass thing seems pretty wrong, but i just want a reliable trigger for it
13:47imirkin_: and i want something i can repro on nvc0 as well, not just nv50 (which has that weirdo pass)
13:48imirkin_: i guess i could force it to do textureGrad() with explicit derivatives and sufficient other conditions to have to hit the manual path...
13:48imirkin_: er hmmmmmmm.... i think i know how to create the control flow i need
13:48tobijk: mh heartstone was mentioned in a bug, that would be a free game
13:48tobijk: maybe you can trigger it with that game?!
13:49tobijk: and in you i mean we :)
13:50imirkin_: well, it also only mysteriously failed on nv50 but worked fine on nvc0 :)
13:51imirkin_: i tried to figure out why, but eventually gave up
13:51imirkin_: i dunno who this guy is, but i can't believe he tracked this down
13:52tobijk: hehe well at least he tries, thats a good start imho
13:57tobijk: ah we have a trace, why did i ignore that
20:04imirkin: skeggsb: do you happen to have notes about vbios opcode 0x59?
20:05imirkin: 0x5a seems simple enough
20:09skeggsb: i've not encountered them so far
20:10imirkin: the function that implements 0x59 is going to be a bit of a pain to work out
20:11skeggsb: what's this for? do we have a binary driver trace of it being used?
20:11imirkin: we have a vbios script that has it
20:12skeggsb: oh wow, i'm rather shocked that it's for something that old, i was thinking this was new
23:05Nebukadneza: yay! with 4.0.0-1 on debian, resume works!
23:05Nebukadneza: thx^^ for the nice driver, freeing me from nvidias shackles!